[1] And even then...
[2] Yes, Colemak
On that page (http://www.newyorker.com/strongbox/, no SSL, no hashing, etc.) is an onion link. I see it as http://tnysbtbxsf356hiy.onion.
An adversary with sufficient access (whether to the server, or to the network between the New Yorker site and your browser) can rewrite that link without your ever knowing. Then you are legitimately directing your browser, once on Tor, to a spoofed site.
To compare, ssh uses no PKI, but is relatively resistant to MITM as it stores public key fingerprints for a given service on first access.
If the Feds (or whoever) were swapping out this address at a large scale, then the NYTs would notice and communicate to the public that this was happening. If the NYT was muzzled and every computer that you or anyone you know has it's connection being tampered with, then you are basically fucked anyway and this is the least of your concerns. At this level of paranoia you shouldn't be trusting the NYT anyway...
If they are not MiTM'ing this at a large scale, then quick verifications on other networks with other peoples' computers will work just fine.
After Manning and some other fun, the military is not really a fan of this happening, but realizes that it would provide an excellent way to identify seditious elements who don't have other channels to work out their moral dilemmas (or just their annoyance at being an expendable cog in the bureaucracy).
MilInTheMiddle proxy rewrites the page, replaces just that link with one that is under their control, then waits. People outside the military see the normal link, but a kid in Iraqistan sees the honeypot. When s/he sends secrets to the New Yorker (it's not even the NYT), the operator has a lead not only on what information may be flowing to the media, but is able to stop it and investigate the sender.
The same case could be made for financial institutions, government agencies, people in foreign countries with national firewalls, etc.
The only reason I'm not particularly concerned about this is that I strongly suspect that a) the organizations don't really have their sh*t together enough to do this in a fashion that wouldn't create immense blowback when (not if) discovered, and b) there are easier methods to get similar information already in place.
Let's reduce this even further. The New Yorker has put up a webpage saying,"Send your secrets here!" Pretend they put up a phone number instead. 1-800-231-2142. The military phone switching system can either automatically redirect calls to that number to a switchboard of impersonators, or they could change the number to one they already have, so that they can also capture military assets using personal phones, mobiles, etc.
The New Yorker page does not give any information about "offline" validation. They give a pointer to a supposedly secure dropbox. That dropbox may actually be secure, but there are a number of ways to subvert it.
Offline verification would work, but if we are assuming extensive MiTMing and meatspace impersonation then offline verification is just something the whistleblower needs to figure out themselves.
"offline verification is just something the whistleblower needs to figure out themselves."
You are seeing the point, albeit as if through a glass, darkly. The New Yorker has not advised anyone of this need, nor have they communicated the additional safeguards one much take, proportional to the risk of interception and identification.
[1] http://en.wikipedia.org/wiki/ECHELON, https://www.eff.org/cases/nsa-multi-district-litigation, black ops, http://en.wikipedia.org/wiki/File:Aldrich_Ames_mailbox.jpg, http://web.archive.org/web/20060210092316/http://www.liberty..., http://books.google.com/books?id=mAR0GI5ggf8C&pg=PA62#v=..., http://www.archives.gov/research/holocaust/finding-aid/civil..., http://www.britannica.com/EBchecked/topic/635080/Sir-Francis..., &c.
You seem to be misunderstanding me. I think that there is value to offline communication with trusted but unanticipated 3rd parties, but I do not think that there is value to the New Yorker pointing this out.
Unless the Tor site is a honeypot. Hmm a bit odd they don't have any offline validation instructions on there ;)
I guess the downloaded source code could always be diffed with a known legit copy, if it was compiled on a trusted machine.
By the way, how does one determine the right amount of paranoia?
The first response to this thread helped me understand the issues better: http://www.wilderssecurity.com/showthread.php?t=228869
- The plainly accessible page needs to be served through HTTPS, preferably secured with an EV-SSL cert. The adversary would then be unable to read or change any of the site's content; i.e. the whistleblower receives the correct onion link.
- The onion site, too, needs to be served through HTTPS. All traffic passing through Tor is encrypted and thus out of reach of the adversary; traffic leaving Tor for the destination site is encrypted yet again and out of reach as well.
Since few CAs would accept to sign certificates for a .onion domain, that site might have to use a self-signed certificate. This solution remains secure for as long as that certificate's fingerprint is posted on the first page and the whistleblower cross-checks the expected and actual fingerprint values.
This should provide reasonable security for the exchanged message contents. In all likelihood, though, the adversary would be able to prove the whistleblower did communicate with the New Yorker.
One high-risk aspect of this whole deal is a possible breach at any of the whistleblower's trusted CAs. Convergence would be of some help there -- assuming the notaries involved are behaving correctly.
*edit: Sorry, I had a bit of a brain fart right there. Traffic to and from the hidden service (the .onion domain) never leave the Tor network and thus remain secure. Skip the second bullet point.